WebLayer is a high level embedding API to support building a browser.
src/content, which is only concerned with a sandboxed multi-process web platform, WebLayer includes modern browser features and Google integration. It's the reusable version of Chrome, which might share some portions of the UI and also its support for all the modern HTML5 and browser features (e.g. UI for permissions, autofill, safe browsing etc...).
While it's built on top of
src/content, the expectation is that the API will hide the Content API.
Mailing list: firstname.lastname@example.org
Bug tracker: Internals>WebLayer
public the C++ and Java public API. These are the only files an app should use
shell sample app
test test harnesses and test-only helper code
tools helper scripts
app internal code which runs at the beginning of each process
browser internal code which runs in the browser process
common internal code which runs in the browser and child processes
renderer internal code which runs in the renderer process
utility internal code which runs in the utility process
If you haven't done this already, you first need to set up an Android build. If you are a Google employee, reach out to email@example.com for internal instructions. Otherwise follow the Android build instructions.
To run the sample app:
$ autoninja -C out/Default run_weblayer_shell $ out/Default/bin/run_weblayer_shell
To run instrumentation tests:
$ autoninja -C out/Default weblayer_instrumentation_test_apk $ out/Default/bin/run_weblayer_instrumentation_test_apk
Note: this may not work on some versions of Android. If you see an error setting the WebView provider when running instrumentation tests, try running the tests using the WebLayer support APK which uses a different loading path:
$ autoninja -C out/Default weblayer_support_instrumentation_test_apk $ out/Default/bin/run_weblayer_support_instrumentation_test_apk
The test script will build and install all necessary APKs.
Make sure you have the following gn arg:
system_webview_package_name = "com.google.android.webview"
The following tests the latest client library against M80:
$ autoninja -C out/Default weblayer_instrumentation_test_versions_apk $ cipd install --root /tmp/M80 chromium/testing/weblayer-x86 m80 $ out/Default/bin/run_weblayer_instrumentation_test_versions_apk \ --test-runner-outdir out/Default \ --client-outdir out/Default \ --implementation-outdir /tmp/M80/out/Release
To run WPT on android against weblayer do the following:
$ export WPT_TEST= test you want to run, relative to wpt directory. $ autoninja -C out/Default run_weblayer_shell weblayer_shell_wpt $ out/Default/bin/run_weblayer_shell $ testing/scripts/run_android_wpt.py --webdriver-binary=out/Default/clang_x64/chromedriver --product android_weblayer --isolated-script-test-output /tmp/weblayer_out.json --include $WPT_TEST --ignore-browser-specific-expectations --ignore-default-expectations
run_android_wpt.py does not install
weblayer-shell, you need to do that yourself (executing
run_weblayer_shell will do that).
To run against clank:
$ export WPT_TEST= test you want to run, relative to wpt directory. $ autoninja -C out/Default monochrome_public_apk $ out/Default/bin/monochrome_public_apk install $ testing/scripts/run_android_wpt.py --webdriver-binary=out/Default/clang_x64/chromedriver --product chrome_android --chrome-package-name org.chromium.chrome --isolated-script-test-output /tmp/clank_out.json --include $WPT_TEST --ignore-browser-specific-expectations --ignore-default-expectations
--ignore-browser-specific-expectations --ignore-default-expectations flags will prevent *Expectations files from being loaded, which helps with error messages on test failures. Once a test is fixed, rerun it without those flags to ensure the Expectations files are correct.
To run against linux with wptrunner (same runner we use on android, which runs normal chrome):
$ export WPT_TEST= test you want to run, relative to wpt directory. $ autoninja -C out/Default wpt_tests_isolate $ cd testing/scripts $ ./run_wpt_tests.py -t Default $WPT_TEST
To run against linux with run_web_tests (same runner we use on CI, which runs content_shell):
$ export WPT_TEST= test you want to run, relative to wpt directory. $ autoninja -C out/Default blink_tests $ ./third_party/blink/tools/run_web_tests.py -t Default external/wpt/$WPT_TEST
-vvvv may be useful if you want to see loads of information about test execution.
A list of known test failures is in
WeblayerWPTExpectations. The values between the brackets at the end of each line list the expected result types that test can have. For example, a test marked as “[ Failure ]” is expected to fail, while a test marked as “[ Failure Pass ]” is expected to be flaky.
Any failing tests should be removed from
WeblayerWPTExpectations file once fixed.
While many WPT tests fail due to features not being implemented in WebLayer, some may fail due to Features that aren‘t getting enabled or switches that aren’t getting passed to the test as they would be for Clank. If a test is failing due to a missing Feature, check the test FieldTrial configuration in
fieldtrial_testing_config.json. A missing switch could have several causes, but the flags that get passed to the test originate from
Telemetry is run against WebLayer, currently on the bot
Telemetry currently only runs on real hardware. Bug 1067712 is for adding support for emulators.
To see the set of stories executed, click on a successful run, search for
performance_weblayer_test_suite and click on the
Googlers can submit jobs against your own patch using pinpoint. At the time of this writing, logcat is not captured for successful runs (1067024). Submitting a pinpoint run against a patch with a CHECK will generate logcat. For such a run, the logcat is viewable by way of: